Modulating complex behaviors with small molecules: good-bye petri dish, hello zebrafish

Drugs that alter neuronal excitability are important for the treatment of psychiatric disorders such as depression, anxiety, and schizophrenia. However, the discovery of neuroactive drugs mainly relied on serendipitous observations of unexpected effects on humans or animals. Hence, few new classes of neuroactive molecules have been discovered in the last half century. Without a profound understanding of psychiatric diseases at the molecular level, researchers often have to employ simple in vitro screening assays. These usually require changes in the activity of a known protein target and therefore are cumbersome to perform as few targets are known. Neither “in vitro” experiments nor cell-based assays are able to account for the complexity of nervous system interactions of entire organisms, and even though cell culture–based screening allows high-throughput procedures, the disadvantage is that drugs that modify nervous system function in new ways will evade detection.

Now, the research groups led by Alex Schier (Harvard University, Cambridge, MA) and Randall Peterson (Massachusetts General Hospital, Harvard Medical School, Charlestown, MA) have developed a fully automated platform for analyzing the behavioral effects of small molecules on embryonic zebrafish.^1,2

Phenotype-based genetic screens in animal models have been extremely fruitful and are unburdened by a priori assumptions about target genes. The same holds true for drug discovery through phenotype-driven screening procedures. The major limitation of this approach, however, is that the molecular targets can be difficult to uncover. One way around this may be to compare phenotypic variables of known compounds with those of novel ones in order to make predictions about their likely modes of action. An important prerequisite for this systems-level approach is that the recorded phenotypes should allow for a great number of easily discernable variation. Surprisingly, zebrafish behavior is one such content-rich readout, even when recorded from very young larvae.

The approach taken by the two teams is similar and relies on a simple procedure. First, a relevant behavior that is variable and content-rich is identified. Second, a high-throughput screening system that automatically records the data is provided. Third, the data are sorted by an algorithm that allows deduction of meaningful corollaries.

David Kokel and colleagues in the Peterson lab shone a strong pulse of light on larvae at 30 hours post-fertilization, which caused them to shake vigorously for a few seconds. ¹ This response was followed by a phase during which the infrequent spontaneous body flexions typical for this developmental stage were suppressed and another light pulse had no effect. They called this stereotypical pattern the photomotor response (PMR), and it was complex enough to record up to 14 quantitative features. Known psychoactive drugs, like the psychostimulant isoproterenol, increased motor activity throughout the PMR, while the anxiolytic diazepam decreased activity. Other substances reproducibly influenced the duration of specific phases. Importantly, the distinct changes elicited by the well-characterized substances suggested that the PMR is controlled by various neurotransmitter pathways and might be suitable for predicting the mode of action for newly discovered drugs. Fourteen thousand small molecules were tested on more than a quarter of a million embryos, resulting in 655 molecules that generated patterns of increased activity and 327 molecules with sedating effects.

The approach taken by Alex Schier's lab was to test the sleep behavior of single 4-day-old zebrafish larvae. ² Following the transition between light and dark, a behavioral fingerprint was recorded for the duration and intensity of bouts of active and resting locomotor activity. Ten percent of more than 5000 different small molecules affected rest/wake activity of the larvae over a duration of 3 days.

The large quantities of data generated by both teams required ingenious ways to extract useful data. To this end, the quantitative features were first translated into a barcode, or fingerprint, representing the specific set of behavioral changes for each small molecule. Next, all barcodes were hierarchically clustered, so that a tree diagram, or dendrogram, could be constructed in which fingerprints of similar behavioral changes clustered closer together than distantly related ones.

These phenoclusters therefore contained functionally similar molecules and could be used to sort substances with different known cellular targets into common pathways that might underlie their similar phenotypes. For example, Peterson's team found that beta-adrenergic receptor agonists were grouped in a cluster of substances that generally elevated activity in all PMR phases, dopamine agonists elicited a longer latency period, and adenosine receptor antagonists were enriched in a cluster with a shorter excitation period.

Several types of correlations became evident. First, substances that elicited similar behavioral profiles often shared annotated targets. Thus, it is possible to suggest likely mechanisms through which poorly characterized substances alter behavior, even if they are structurally unrelated. Two novel acetylcholinesterase inhibitors, whose activities could be reproduced in in vitro tests, were identified based upon this principle. Second, classes of drugs with a shared molecular structure produced correlated behaviors. When dendrograms were constructed that were not based on phenotypic similarity, but on similarity of compound structure, chemical variations of a common molecular backbone emerged that caused similar phenotypes. These might provide starting points for exploring potential variations of compounds with improved properties. Third, drugs that act on more than one target clustered with drugs that share the same set of targets. And fourth, substances in zebrafish frequently induced effects similar to those seen in mammals, although the authors point out that they found differences as well in the response to psychotropic drugs between zebrafish and humans.

Of course this research also generated new insights into animal behavior. Schier's team, for example, found that anti-inflammatory compounds, which are known to help patients sleep during an infection, also increased waking activity during the day. Another finding was that the regulation of total rest, the process of falling asleep, and waking activity are disassociated and probably controlled by distinct mechanisms.

Automated high-throughput behavioral screens may not be limited to screening healthy animals, but could also be used to screen for substances that revert abnormal phenotypes resulting from genetic changes or pharmacological treatments. An intriguing possibility is to use genetically altered fish, through knockout or transgenic methods, that are designed to mimic diseases of the nervous system and for which new compounds are urgently needed.

The gape of the icefish

Icefish are a very peculiar type of animal. First, they have evolved antifreeze proteins that allow them to thrive in the cold temperatures of the Antarctic waters. Second, because cold water carries more oxygen, they can afford to have drastically reduced numbers of hemoglobin molecules, leaving some species with colorless blood. Thirty million years ago, the Antarctic ecosystem changed dramatically when the waters between South America's southern tip and Antarctica opened up, leading to the formation of a thermally isolated circumpolar water current. In its wake, temperatures around the arctic circle started to drop.

Today's icefish (Notothenioids) evolved between 14 and 10 million years ago from perches that had lost their swim bladders owing to their lifestyle on the ocean's floor. Like today's benthic populations, they had solid and heavy jaws with several rows of large teeth. Their expansion into new habitats, which included a pelagic lifestyle in the open Antarctic seas, meant that they had to regain control over buoyancy. The evolutionary loss of the swim bladder was irreversible, but icefish developed alternative ways to reduce body density. The successful solution was based on lighter skeletons and reduced mineralization rate of skeletal tissue.

Another adaptation toward a more efficient hunting style in open water was the evolution of a wider jaw gape and a more streamlined body shape. True pelagic species have short jaws with a few large oral teeth, ideally suited to suck in small invertebrates. Other species that engulf shoals of small fish in their mouths in a forward movement, called ram-feeding, have evolved longer jaws with many small teeth to capture their prey more efficiently.

The evolutionary trend toward reduction of heavy and solid jaws in pelagic icefish and the shortening of jaw bones is achieved by pedomorphosis (i.e. the retention of juvenile features in adult organisms). Craig Albertson (Syracuse University, Syracuse, NY) and colleagues from Italy and the United States have begun to unravel the molecular basis of delayed development in pelagic icefish (Albertson et al., 2010). They compared the rates of mineralization in the different ecotypes and determined the onset and duration of marker gene expression for cartilage and bone.

The jaw and branchial skeleton are first formed as cartilage, which is then gradually replaced by bone material. Pelagic fish were found to have a prolonged cartilage stage (examined through expression of the cartilage-specific collagen gene col2a1) and a much later onset of markers for osteogenic cells (col1a1 and col10a1). In contrast, the bones of the pectoral fins formed at the same stages in benthic and pelagic forms, showing that pedomorphosis acted locally on some bones only.

The temporal sequence of head skeleton formation in icefish also differed from that in outgroups. While jaw bones in stickleback and zebrafish develop in a more or less anterior to posterior fashion, icefish developed ventral and anterior bones first and failed to elaborate certain dorsal elements.

Unexpectedly, development of the pharyngeal jaws, a second set of jaws in the back of the pharynx, was severely delayed and in pelagic hunters formed as virtually the last set of head bones. This is quite unusual, as pharyngeal jaws are important for crushing the shells of crustacean prey and are among the first skeletal elements to mineralize in zebrafish. The authors suggest that delayed pharyngeal jaw formation might not actually be adaptive, but could be the result of a developmental compromise in which certain processes cannot be dissociated from others. This developmental constraint limits the degree to which evolution can shape certain parts of the head skeleton without affecting others. More work on these fascinating fish is highly desirable to unravel how heterochronic shifts in developmental pathways can influence the evolution of skeletal development and tissue mineralization and to determine how icefish lineages were able to radiate despite high rates of gene flow between populations that are large distances apart.

Comparing induced pluripotent stem cells to the regeneration blastema: apples and oranges or half the road to de-differentiation?

Our knowledge of what it takes to reprogram differentiated somatic cells towards pluripotent stem cells has dramatically advanced during the past couple of years. The key to why this process is possible at all was the finding that a suite of genes, so-called pluripotency genes, when expressed in somatic cells, can initiate a de-differentiation program that results in induced pluripotent stem (iPS) cells. These resemble natural pluripotent stem cells in many ways and can differentiate into a variety of other cell types. By now, a number of pluripotency genes are known, and while a few are sufficient to induce pluripotency, a host of others promote self-renewal of iPS cells.

Reprogramming of differentiated cells is a phenomenon that generations of scientists have tried to understand. Fish and amphibians possess the amazing capacity to regenerate a variety of organs upon amputation. During this process of epimorphic regeneration, a new tissue type, the blastema, forms from cells derived from the remaining tissue. Recent work on the regenerating limb of the axolotl (a salamander) had suggested that blastemal cells are restricted pluripotent cells that give rise mostly to the same cell type as before. Rather than being a single cell type, the blastema appears to be a mix of different cell lineages, each with a restricted differentiation outcome. Thus the time was ripe to test whether a common mechanism regulates the processes of epimorphic regeneration and the induction of pluripotent stem cells. Bea Christen, Vanesa Robles, and colleagues in Juan Carlos Izpisúa-Belmonte's lab (Center for Regenerative Medicine of Barcelona, Spain) tested this hypothesis by asking whether pluripotency markers are up-regulated during regeneration in zebrafish caudal fins and limbs and tails of Xenopus tadpoles (Christen et al., 2010).

Four transcription factors, Oct-4, Sox2, Klf4, and c-Myc, are sufficient to revert differentiated somatic cells into iPS cells. Izpisúa-Belmonte's team used quantitative real time PCR to determine whether the expression of these and a number of self-renewal markers was enhanced during regeneration. Xenopus animal cap cells, which served as examples for embryonic pluripotent cells, expressed all but two markers. In contrast, the tadpole blastema cells expressed barely half of 11 tested pluripotency genes, none of which was up-regulated during regeneration. Furthermore, these markers were not specific to regenerating tissue and were also found in regeneration-incompetent limbs of metamorphosing tadpoles. Ultimately, only three factors were found to be expressed at similar or higher levels than in iPS cells. Furthermore, if blastemal cells are truly pluripotent, they should express pluripotency-associated genes uniformly across the regenerate. Reprogramming factors did not conform to this pattern either: of the three most strongly expressed markers, one was down-regulated in all blastemal cells and another one was asymmetrically expressed in the posterior portion of the limb blastema. Last, while their faster cell cycles are another hallmark of pluripotent cells, blastema cells turned out to be cycling more slowly, again indicating that they behave rather like somatic cells.

In zebrafish, it turned out that most pluripotency genes are also expressed in nonregenerating fins and none are up-regulated in the blastema. The authors then wondered whether reprogramming factors are required for regeneration. They blocked the activities of the oct-4 ortholog pou5f1, as well as sox2, both of which are key factors in self-renewing pluripotent cells, by injecting morpholinos into blastemas at 1, 2, or 3 days after amputation. Both factors appeared to slow down the regrowth of amputated tissue, and while pou5f1 seemed to be required at all tested time points, sox2 knockdown only affected the late regenerative outgrowth stages when cells start to re-differentiate.

Taken together, the findings harmonize with the observed lineage restriction of blastema cells in regenerating axolotl limbs and suggest that blastema cells and pluripotent cells have little in common. At best, the blastema is a multipotent tissue that expresses low levels of some or all key pluripotency factors. One appealing hypothesis emerging from this study is that these low levels might start reprogramming but then stall at an early stage. In this view, multipotency not only is the best possible result a blastema can achieve, but appears like a more economical response to injury than full reversal to pluripotency.